jsp@milton.u.washington.edu (Jeff Prothero) (08/04/90)
A recent discussion here focussed on the difficulty of building nonrotating optical mass storage due to difficulty of steering the beam. Why not use electron beams to read/write nonrotating electric mass storage? Points in favor: o Established technology -- used in some of the earliest computers :-). More seriously, CRT tubes *are* a well-developed technology in general. o Potential storage density higher than optical. (We use scanning electron microscopes to see features too small for optical microscopes.) o Potential low seek times. How long does it take a really fast oscilliscope to seek side- to-side? o Potential non-volatile storage. EEPROMs and electrets show that electric charge *can* be stored for considerable periods. Is there some fundamental reason electron-beam mass storage can't compete with optical and magnetic mass storage? Has any work been done on this since the 50s?
lindsay@MATHOM.GANDALF.CS.CMU.EDU (Donald Lindsay) (08/04/90)
In article <JSP.90Aug3123600@milton.u.washington.edu> jsp@milton.u.washington.edu (Jeff Prothero) writes: >Why not use electron beams to >read/write nonrotating electric mass storage? >Has any work been done >on this since the 50s? Yes, there has been work on this in the last decade or so. >o Potential storage density higher than optical. > (We use scanning electron microscopes to see > features too small for optical microscopes.) Electron beams can indeed be small. However, the charge has to be stored. The last proposal I remember was going to fabricate capacitors onto a wafer. This differs from a DRAM in that the beam replaces some or all of the wiring pattern. Also, this is wafer- scale, of course. I'm not sure what happened. Perhaps the fabrication people refused to stop improving DRAM. Perhaps it was the bulk of the tube. Maybe it was the lack of parallelism - one beam means one bit wide, no? Or maybe the projects are still going on. I know that some things, like CCD memory, died in spite of being viable. ( CCD was intrinsically about twice as dense as DRAM, but a factor of two is just a one-year phase angle in this business.) -- Don D.C.Lindsay
elg@cypress.UUCP ( Marketing) (08/05/90)
Electron beam addressable memory has been worked on since the 50's. The Magnetic Peripherals Inc. subsidiary of Control Data Corp. (now Imprimus, a subsidiary of Segate) had a working design in the late 70's. There were multiple problems with storage tube arcing and relative density to magnetic media, but it was fast. I believe each storage tube had an addressable matrix of 1024x1024 bits. The controller supported multiple tubes. The acronym for the machine was EBAM and it used a controller I believe based on the old CDC 5600 MPP. This controller was quite fast for its day but would be replaced by a small handfull of chips today. Ed Grivna Cypress Semiconductor
lm@snafu.Sun.COM (Larry McVoy) (08/05/90)
In article <JSP.90Aug3123600@milton.u.washington.edu> jsp@milton.u.washington.edu (Jeff Prothero) writes: >Is there some fundamental reason electron-beam >mass storage can't compete with optical and >magnetic mass storage? Has any work been done >on this since the 50s? CRT's produce a fair amount of radiation. I'd just as soon not add to the amount of zapping I get. --- Larry McVoy, Sun Microsystems (415) 336-7627 ...!sun!lm or lm@sun.com